@zekjur@mas.to @siracusa@mastodon.social fascinating read ! To the Appendix B, AI section… is there a reason why you didn’t use Mistral AI’s models in your tests ? I know that generally they’re not as good as OpenAI’s/Anthropic’s, but being the only non-US/non-China company that could claim to have a frontier model (debatable, but it does come close), I think it’d be nice to include in the data.